CN113176211A - Day cloud detection method based on HSV color space model transformation - Google Patents

Day cloud detection method based on HSV color space model transformation Download PDF

Info

Publication number
CN113176211A
CN113176211A CN202110458849.3A CN202110458849A CN113176211A CN 113176211 A CN113176211 A CN 113176211A CN 202110458849 A CN202110458849 A CN 202110458849A CN 113176211 A CN113176211 A CN 113176211A
Authority
CN
China
Prior art keywords
cloud
color space
hsv color
space model
cloud detection
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110458849.3A
Other languages
Chinese (zh)
Other versions
CN113176211B (en
Inventor
李倩媚
陈楚群
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China Sea Institute of Oceanology of CAS
Original Assignee
South China Sea Institute of Oceanology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by South China Sea Institute of Oceanology of CAS filed Critical South China Sea Institute of Oceanology of CAS
Priority to CN202110458849.3A priority Critical patent/CN113176211B/en
Publication of CN113176211A publication Critical patent/CN113176211A/en
Application granted granted Critical
Publication of CN113176211B publication Critical patent/CN113176211B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/55—Specular reflectivity
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1793—Remote sensing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a daytime cloud detection method based on HSV color space model transformation, which comprises the steps of carrying out radiometric calibration on acquired remote sensing data to obtain reflectivity information of a target waveband; analyzing the reflection characteristics of four objects, namely thick cloud, thin cloud, seawater and land, in a preset waveband or waveband combination according to the reflectivity information, selecting a plurality of wavebands or waveband combinations reflecting cloud information, and sequentially putting the wavebands or waveband combinations into a red channel, a green channel and a blue channel to generate an RGB false color image; calculating to obtain chromaticity, saturation and brightness components according to an HSV color space model conversion equation; constructing a daytime cloud detection index algorithm NDCD according to the distribution characteristics of cloud pixels in the calculated component dataday(ii) a NDCD is calculated by utilizing daytime cloud detection index algorithmdayAnd (5) counting the histogram, setting a threshold value according to the distribution characteristics in the histogram and detecting the polluted pixels. The NDCD of the inventiondayThe algorithm effectively reduces the thinnessAnd the cloud misjudgment and missed judgment conditions have higher reliability, universality and accuracy.

Description

Day cloud detection method based on HSV color space model transformation
Background
A layer of time-thin and time-dense thick cloud layer is formed on the earth surface, is a visible polymer which is formed by small water drops formed by liquefaction of water vapor when meeting cold or small ice crystals formed by sublimation and floats in the air, is a natural phenomenon of the earth and influences the landing surface radiation balance and the energy balance. At each instant, about 50% of the earth's surface is covered by various clouds. The cloud has great influence on solar radiation received by the ground object and ground object thermal radiation received by the sensor, and in the process of acquiring the real SST (sea Surface temperature) by using the thermal infrared remote sensing technology, the cloud weakens the real record of the sensor on the sea Surface radiation value by blocking the sea Surface, so that the inversion result of the pixel SST polluted by the cloud is low, and the cloud detection becomes one of key technologies for satellite remote sensing image preprocessing. Meanwhile, in the quantitative research work of the marine elements, an accurate cloud detection step is also indispensable.
In a visible light channel of the remote sensing image, the reflectivity of the cloud is obviously higher than that of the water body. In the infrared channel, the cloud top brightness temperature is obviously lower than the water body temperature, so that a good cloud detection result can be obtained by setting a threshold value for the visible light and the infrared channel. For the ocean underlying surface, the background is not complex, but the reflectivity and the brightness temperature value of the thin cloud are close to those of a water body, and the thin cloud pixel is easy to be missed, so that the remote sensing quantitative inversion result of ocean physical parameters is influenced, and larger deviation is caused. Therefore, how to accurately and quickly identify the thin cloud and the small volume rolling cloud is a difficult point of the current cloud detection work.
For decades, a great deal of research work is done by predecessors on the aspect of cloud detection of remote sensing data, various automatic cloud detection algorithms are proposed, and accurate judgment of cloud pollution pixels is expected to be achieved, so that SST inversion accuracy is improved. Rosslow et al proposed an ISCP (International software Cloud computing project) Cloud detection algorithm based on a visible light narrow band of 0.6 μm and a thermal infrared band of 11 μm in 1989, in which the difference between the radiation value of each pixel and the clear sky radiation value is greater than the change of the clear sky radiation value itself as a determination threshold of the Cloud pixel, and the uncertainty of the detection result is excessively influenced by the threshold. In 1991, CLAVR (CLoud from AVhrR) algorithm is established based on five channels of AVHRR (advanced virtual high resolution radiometer) sensor, such as Srown, and the like, space difference is used as a judgment condition, a 2 x 2 pixel matrix is used as a detection window, and the pixel matrix is judged according to the judgment results of 4 pixel points in the windowThe algorithm is not efficient due to the large number of cycles. Wylie et al, 1994, proposed CO based on HIRS multi-spectral data2Flake method using CO2The absorption wave band detects clouds of different levels, the detection on the thin roll cloud is effective, but when the difference between a cloud pixel and a clear space pixel is smaller than the noise of an instrument, an algorithm is invalid. Wisetphonickhikij and Dejhan use wavelet analysis method to perform cloud detection in 1999, and fuse multi-resolution wavelet decomposition images to achieve the purpose of eliminating cloud pollution, but the method is complex in calculation. Azimi-Sadja and Zekavat in 2000 introduce a machine learning algorithm Support Vector Machine (SVM) (support Vector machines) into the field of cloud detection, the method needs to standardize the feature of the ground feature, select samples for training and learning, and the detection effect depends on the selection of a classification method and the representativeness of the feature samples.
The research of domestic cloud detection starts relatively late, a cloud system feature library is established based on 177 GMS infrared cloud pictures in 2001, champion fragrance and the like, then automatic segmentation is carried out by utilizing a neural network, and the feasibility of judging cloud pixels by combining multiple thresholds and an artificial neural network is proved. In 2003, in Song Xiaoning and Zhao Ying, according to the spectral characteristics and texture characteristics of the cloud pixels, a cloud automatic detection algorithm based on space structure analysis and a neural network is provided, and multiple cloud detection algorithms are subjected to comparison analysis and mutual verification. In 2007, Mafang and the like design an infrared split window channel difference value cloud detection algorithm according to the statistical characteristics of GMS-5 satellite cloud pictures in each channel, and can be used under the condition of any solar altitude angle by combining with a water vapor channel threshold value. According to satellite data such as FY-2C, MTSAT and the like, in 2010, Guochong and the like, difficulty in selecting threshold values of different underlying surfaces is eliminated through a multi-channel space-time characteristic comparison test, and an optimized ISCCP cloud detection algorithm is provided. The method is characterized in that the whole army in 2011 establishes a multi-spectral threshold daytime cloud detection algorithm based on FY-3A/VIRR data and on five underlying surfaces of ice, snow, desert, coast, land and water. In 2011, the HSV color space is introduced into the field of cloud detection for the first time by Li micro and Li Deren, and an MODIS cloud detection algorithm is provided, and has the characteristics of simplicity, high timeliness, high precision and the like, but the situation of thin clouds is not considered. 2016, the junk population establishes a BP (Back propagation) neural network algorithm based on MODIS data, performs cloud detection and cloud phase recognition by using the algorithm, and finally compares the algorithm with an MOD06 product to find that the algorithm is quick and accurate and has strong autonomous learning capability. Zhengrong and the like in 2019 are based on textural features of remote sensing images, and a rapid cloud detection algorithm is established by using an improved maximum response filter and a K-means clustering analysis method.
Although the methods are numerous, most of the methods have the defects of poor adaptability, low detection efficiency, strong subjectivity of results and the like. Meanwhile, if the algorithms are directly transplanted into a preprocessing flow of domestic satellite data, missing judgment and misjudgment of cloud pixels can be caused due to limitations of the algorithms, and SST inversion errors can be further caused.
Disclosure of Invention
The invention provides a daytime cloud detection method based on HSV color space model transformation, which aims to overcome the defects of poor adaptability, low detection efficiency, strong subjectivity of results and the like and can better identify thin clouds and acquire more accurate SST inversion data.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a daytime cloud detection method based on HSV color space model transformation, which comprises the following steps:
carrying out radiometric calibration on the acquired remote sensing data to obtain reflectivity information of a target waveband;
analyzing the reflection characteristics of four objects, namely thick cloud, thin cloud, seawater and land, in a preset waveband or waveband combination according to the obtained reflectivity information, selecting a plurality of wavebands or waveband combinations reflecting cloud information, and sequentially putting the wavebands or waveband combinations into a red channel, a green channel and a blue channel to generate an RGB false color image; the thick cloud is a cloud layer which can not be penetrated by more than 75% of visible light in the pixel, and other cloud layers are thin clouds;
calculating the generated RGB false color image according to an HSV color space model conversion equation to obtain chromaticity, saturation and brightness components;
based on the calculated chroma, saturation and brightness degree measures and the number of brightness componentsConstructing a daytime cloud detection index algorithm NDCD according to the distribution characteristics of cloud pixels in the cloud imageday；
NDCD is calculated by utilizing daytime cloud detection index algorithmdayAnd (5) counting the histogram, setting a threshold value according to the distribution characteristics in the histogram and detecting the polluted pixels.
Further, the remote sensing data is FY-4A/AGRI L1 grade remote sensing data, and the conversion of a digital quantization value and an apparent reflectivity is completed through the following linear relation according to a radiometric calibration coefficient contained in an FY-4A/AGRI L1 data file:
R＝Scale×DN+Offset
where DN represents the digital quantized value recorded by the sensor, stored as a binary scientific data set, Scale represents the gain value, and Offset represents the Offset value.
Further, the reflectivity information of the target wavelength band includes: reflectance information of visible light in the 0.47 μm, 0.65 μm band, near infrared in the 0.83 μm, 1.38 μm band, and short wave infrared in the 1.61 μm, 2.23 μm band.
Further, according to the reflectivity characteristics of the four objects of thick cloud, thin cloud, seawater and land in the visible light, near infrared and short wave infrared bands of the FY-4A/AGRI sensor, the wave band or the wave band combination capable of clearly displaying the four object pixels is selected, the reflectivity of the four objects is normalized, the characteristic distribution of the four objects is analyzed, and the R is selected0.83、R0.65And R0.83/R0.65And sequentially putting the red, green and blue channels to generate an RGB false color image.
Further, the calculation formula of the normalization operation is as follows:
Figure RE-GDA0003097185470000031
in the formula, XnormRepresenting data obtained by normalisation calculation, X representing the original data of the band or band ratio, Xmax、XminRepresenting the maximum and minimum values in the original data set, respectively.
Further, the chrominance, saturation and luminance components calculated according to the HSV color space model conversion equation are:
Figure RE-GDA0003097185470000041
Figure RE-GDA0003097185470000042
Figure RE-GDA0003097185470000043
in the formula, R, G, B represents red, green, and blue components in an RGB color space, H, S, V represents hue, saturation, and brightness components in an HSV color space, and max is max (R, G, B) and min is min (R, G, B).
Further, a daytime cloud detection index algorithm NDCDdaComprises the following steps:
Figure RE-GDA0003097185470000044
in the formula of NDCDdayFor daytime cloud detection index, H, V are hue and brightness components in the HSV color space model, respectively.
Further, the cloud detection index algorithm NDCDdayThe threshold value of (a) is obtained by:
NDCD obtained through cloud detection index algorithm calculationdayAnd a spatial distribution diagram and a statistical histogram thereof, wherein the histogram comprises four main peak values of seawater, land, thick cloud and thin cloud, and the NDCD corresponds to valley values connected among the peak values of each objectdayThe index value sets a threshold value.
Compared with the prior art, the invention has the beneficial effects that:
the method finds out the most reflective clouds by the reflection characteristics of four objects of seawater, land, thick cloud and thin cloud in each wave band in the daytimeThe three wave bands or wave band combinations of the information generate RGB false color images, chrominance and luminance component data are obtained by utilizing HSV color space model transformation, and a cloud detection index algorithm is constructed. Meanwhile, the detection threshold is set according to the conditions of different images through the statistical histogram, so that the identification result of the thin cloud is more accurate. Compared with the cloud detection product of the national satellite center, the NDCD provided by the inventiondayThe algorithm effectively reduces the misjudgment and the missed judgment of the thin cloud, and has higher reliability, universality and accuracy.
Drawings
Fig. 1 is a flowchart of a daytime cloud detection method based on HSV color space model transformation according to embodiment 1 of the present invention;
FIG. 2 is a schematic diagram of an RBG and HSV color space model;
FIG. 3 is a line graph of reflection characteristics of seawater, land, thick clouds and thin clouds in the visible, near infrared and short wave infrared bands of the FY-4A/AGRI sensor;
FIG. 4 is a cloud detection index distribution and statistical histogram thereof;
FIG. 5 is an NDCDdayA comparison graph of cloud detection algorithm results and cloud detection products of a national satellite center;
FIG. 6 is X in FIG. 51And X2Enlargement of the area and its corresponding RGB near-true color image (R)0.83、R0.65、R0.47Putting red, green and blue channels in turn).
Detailed Description
The invention will be further described with reference to the accompanying drawings and the detailed description below:
example (b):
at present, the mainstream cloud detection algorithm is a multiband threshold method, although the implementation is easy, the steps are relatively complex, multiple band operations need to be explored and researched one by one to give corresponding thresholds, the detection results can be obtained only by integrating the thresholds, and then the detection results need to be returned to carry out multiple times of debugging to find the appropriate thresholds.
In order to more efficiently and accurately judge the cloud pollution pixels, the cloud detection algorithm with higher efficiency, applicability and accuracy is constructed by thinking from the image processing direction according to the spectral characteristics of each wave band of FY-4A/AGRI primary (L1B) data and the spectral reflection and radiation characteristics of each object.
The HSV color space model (shown on the right of figure 2) is more frequently used in image processing, because the HSV color space model is closer to the perception experience of people on color than the RGB model (shown on the left of figure 2), the actual color can be better explained, and the hue, the vividness and the brightness of the color can be more intuitively expressed. The model is an inverted cone that can be represented by a coordinate system in three-dimensional space. Where the cone height represents the color intensity V, the top surface represents the brightest color (i.e., V ═ 1), and the cone apex at the bottom represents the darkest black color (i.e., V ═ 0), it is noted that there is no direct connection between the intensity value and the intensity of the light. The angle of rotation about the V axis represents hue H, ranging from 0 to 360 °, and when the angle is 0 °, 120 °, 240 °, it represents red, green, blue, respectively, each color differing from their complementary colors by 180 °. The distance from the center of the top surface to the circumference represents the saturation S, and the radius of the top surface of the cone is 1 because the value range of the saturation is 0-1.
The HSV space transformation method can be based on the spectral difference of cloud and sea surface radiation as a theoretical basis, is based on the positive transformation of a color space into a mathematical basis, and has the characteristics of high calculation speed, simplicity, feasibility, high precision, suitability for different seasons and the like, so that the invention considers the method as the basis, and establishes an efficient cloud detection algorithm by carrying out comparative analysis on different wave band combinations.
The daytime cloud detection process based on the HSV color space transformation model is realized by adopting the method disclosed by the invention and is shown in figure 1, and the method is described in detail by combining the accompanying drawings and the specific implementation mode:
101. the method comprises the following steps of utilizing self gain and offset coefficients in an FY-4A/AGRI L1 level remote sensing data file to conduct radiometric calibration on FY-4A/AGRI L1 level remote sensing data, and completing conversion of DN values and apparent reflectivity through the following linear relational expression:
R＝Scale×DN+Offset
where DN represents the digital quantized value recorded by the sensor, stored as a binary scientific data set, Scale represents the gain value (slope) and Offset represents the Offset value (intercept). Apparent reflectivity data can be obtained after each wave band is subjected to radiometric calibration, and the value range is 0-1.
102. The reflection characteristics of four objects of thick cloud, thin cloud, seawater and land in each wave band or wave band combination are analyzed firstly. As can be seen from fig. 3, the cloud has a very high reflectance (greater than 60%) in the 0.65 μm visible band. The reflectivity of some clouds can reach more than 70 percent. While the reflectivity of the water body in the band is low (less than 10 percent), and the reflectivity of the thin cloud and the land is centered. Therefore, the reflectivity (R) of the 0.65 μm wave band0.65) Can easily distinguish the water body pixels. Meanwhile, the water body has strong absorptivity to the near-infrared band of 0.83 μm, and the surface reflectivity of the water body without solar flare is lower than 5%. The thick cloud, the thin cloud and the land still have relatively high reflectivity, so that the 0.83 mu m wave band also has better capability of identifying the water body pixels. This band reflectivity (R) is used in the Suomi-NPP/VIIRS Cloud mask algorithm VCM (VIIRS Cloud mask)0.83) As a basis for identifying the water body pixels in the daytime. Also as shown in fig. 3, the reflectance of both the thick cloud and the land at the 0.83 μm band is higher than that of the 0.65 μm band, the thin cloud has similar reflectance in both bands, and the reflectance of only the water body at the 0.83 μm band is lower than that of the 0.65 μm band. Therefore, the ratio of the reflectivity of the 0.83 μm band to the reflectivity of the 0.65 μm band (note R)0.83/R0.65) The water body picture elements can be identified more efficiently. Then R is0.83、R0.65And R0.83/R0.65The three wave bands or wave band combinations which can reflect the cloud information most are sequentially placed into a red channel, a green channel and a blue channel to generate RGB false color images, wherein the land presents purple, the seawater presents black, the thick cloud presents bright yellow, and the thin cloud presents light yellow.
In the process of carrying out comparison evaluation on the wave band combinations, in order to enable the wave band combinations to be in the same order, normalization calculation is carried out, namely data of wave band reflectivity or wave band ratio are linearly stretched to be in a [01] range. The normalized calculation equation is as follows:
Figure RE-GDA0003097185470000061
in the formula, XnormRepresenting data obtained by normalisation calculation, X representing the original data of the band or band ratio, Xmax、XminRepresenting the maximum and minimum values in the original data set, respectively.
103. Obtaining an HSV color image and corresponding three data of chroma H, saturation S and brightness V according to the following HSV color space model conversion equation:
Figure RE-GDA0003097185470000071
Figure RE-GDA0003097185470000072
Figure RE-GDA0003097185470000073
in the formula, R, G, B represents red, green, and blue components in an RGB color space, H, S, V represents hue, saturation, and brightness components in an HSV color space, and max is max (R, G, B) and min is min (R, G, B). Wherein the chromaticity represents the color of an object, and the sea water, the land, the thick cloud and the thin cloud respectively represent different colors. The saturation degree indicates the purity of the color, and the value thereof depends on the ratio of the color-containing component and the achromatic component (gray). The brightness represents the physical quantity of the reflection (or luminescence) intensity of the surface of the reflector (or luminophor), and here represents the ratio of the four objects to reflect solar radiation and the ratio of the four objects to receive the solar radiation, and each object has a specific ratio range in different wave bands.
104. Constructing a daytime cloud detection algorithm NDCD according to the distribution characteristics of cloud pixels in the component dataday. Through the chrominance and luminance components, four objects can be easily distinguished,the object can not be distinguished through the saturation S, so that the cloud detection index equation is constructed by utilizing the normalized data of the chromaticity H and the brightness V. Wherein the chromaticity of seawater is similar to that of land, thin clouds, and its brightness is the smallest of the four objects. In order to obtain a better cloud detection index equation, three combined calculation forms of chromaticity and brightness are discussed, and finally the cloud detection index equation is determined to be in the following form:
Figure RE-GDA0003097185470000074
in the formula of NDCDdayFor daytime cloud detection index, H, V are hue and brightness components in the HSV color space model, respectively.
105. NDCD obtained through cloud detection index algorithm calculationday(range-1-1) and its spatial distribution and statistical histogram (see FIG. 4) with four major peaks of seawater, land, thick clouds and thin clouds, according to the NDCD corresponding to the connected valleys between the peaks of each objectdayAnd setting a threshold value for the index value, so that the identification of the day cloud pixel can be realized (the detected cloud pixel is marked as 1, and if not, the detected cloud pixel is 0, and the result is shown in fig. 5 (a)). By utilizing the cloud detection result, clear-sky seawater remote sensing data can be obtained, and purer and more accurate input parameters are provided for subsequent scientific researches such as quantitative inversion of ocean parameters, monitoring of sea-air changes and the like.
The applicability analysis of the invention to the FY-4A/AGRI remote sensing image is as follows:
taking multi-scene FY-4A/AGRI images at different dates and different moments as application examples, comparing and analyzing the cloud detection result with the corresponding cloud detection product and RGB near-true color image of the national satellite center (see fig. 5 and 6), and finding that the conditions of missed judgment and false judgment of the thin cloud in the cloud detection product of the satellite center are NDCDdayThe cloud detection result is effectively improved. The cloud detection index algorithm established by the invention is simple, convenient and quick in steps and more accurate in thin cloud identification, can furthest judge and identify all cloud-polluted pixels in an image, and verifies NDCDdayThe reliability and accuracy of the algorithm.
Various other modifications and changes may be made by those skilled in the art based on the above-described technical solutions and concepts, and all such modifications and changes should fall within the scope of the claims of the present invention.

Claims (8)

1. A daytime cloud detection method based on HSV color space model transformation is characterized by comprising the following steps:
carrying out radiometric calibration on the acquired remote sensing data to obtain reflectivity information of a target waveband;
analyzing the reflection characteristics of four objects, namely thick cloud, thin cloud, seawater and land, in a preset waveband or waveband combination according to the obtained reflectivity information, selecting a plurality of wavebands or waveband combinations reflecting cloud information, and sequentially putting the wavebands or waveband combinations into a red channel, a green channel and a blue channel to generate an RGB false color image; the thick cloud is a cloud layer which can not be penetrated by more than 75% of visible light in the pixel, and other cloud layers are thin clouds;
calculating the generated RGB false color image according to an HSV color space model conversion equation to obtain chromaticity, saturation and brightness components;
constructing a daytime cloud detection index algorithm NDCD according to the calculated chromaticity, saturation and brightness degree measurement values and the distribution characteristics of cloud pixels in the brightness component dataday；
NDCD is calculated by utilizing daytime cloud detection index algorithmdayAnd (5) counting the histogram, setting a threshold value according to the distribution characteristics in the histogram and detecting the polluted pixels.
2. The method of claim 1, wherein the method of daytime cloud detection based on HSV color space model transformation,
the remote sensing data is FY-4A/AGRI L1 grade remote sensing data, and the conversion of a digital quantization value and an apparent reflectivity is completed through the following linear relational expression according to a radiometric calibration coefficient contained in an FY-4A/AGRI L1 data file:
R＝Scale×DN+Offset
where DN represents the digital quantized value recorded by the sensor, stored as a binary scientific data set, Scale represents the gain value, and Offset represents the Offset value.
3. The HSV color space model transform-based daytime cloud detection method of claim 2, wherein the reflectivity information for the target wavelength band comprises: reflectance information of visible light in the 0.47 μm, 0.65 μm band, near infrared in the 0.83 μm, 1.38 μm band, and short wave infrared in the 1.61 μm, 2.23 μm band.
4. The method of claim 3, wherein based on the reflectivity characteristics of the thick cloud, thin cloud, sea water and land four objects in the visible light, near infrared and short wave infrared bands of the FY-4A/AGRI sensor, the bands or band combinations capable of clearly displaying the pixels of the four objects are selected, the reflectivity of the four objects is normalized, and R is selected by analyzing the characteristic distribution of the four objects0.83、R0.65And R0.83/R0.65And sequentially putting the red, green and blue channels to generate an RGB false color image.
5. The HSV color space model transform-based daytime cloud detection method of claim 4, wherein the normalization operation is calculated by the formula:
Figure FDA0003041604060000011
in the formula, XnormRepresenting data obtained by normalisation calculation, X representing the original data of the band or band ratio, Xmax、XminRepresenting the maximum and minimum values in the original data set, respectively.
6. The HSV color space model transformation-based daytime cloud detection method of claim 5, wherein the hue, saturation and brightness components calculated according to the HSV color space model transformation equation are:
Figure FDA0003041604060000021
Figure FDA0003041604060000022
Figure FDA0003041604060000023
in the formula, R, G, B represents red, green, and blue components in an RGB color space, H, S, V represents hue, saturation, and brightness components in an HSV color space, and max is max (R, G, B) and min is min (R, G, B).
7. The HSV color space model transform-based daytime cloud detection method of claim 6, wherein a daytime cloud detection index algorithm NDCDdaComprises the following steps:
Figure FDA0003041604060000024
in the formula of NDCDdayFor daytime cloud detection index, H, V are hue and brightness components in the HSV color space model, respectively.
8. The HSV color space model transform-based daytime cloud detection method of claim 7, wherein the cloud detection index algorithm NDCDdayThe threshold value of (a) is obtained by:
NDCD obtained through cloud detection index algorithm calculationdayAnd a spatial distribution diagram and a statistical histogram thereof, wherein the histogram comprises four main peak values of seawater, land, thick cloud and thin cloud, and the peak values are connected according to the connection among the peak values of all objectsNDCD corresponding to the valley value ofdayThe index value sets a threshold value.
CN202110458849.3A 2021-04-27 2021-04-27 Day cloud detection method based on HSV color space model transformation Active CN113176211B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110458849.3A CN113176211B (en) 2021-04-27 2021-04-27 Day cloud detection method based on HSV color space model transformation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110458849.3A CN113176211B (en) 2021-04-27 2021-04-27 Day cloud detection method based on HSV color space model transformation

Publications (2)

Publication Number Publication Date
CN113176211A true CN113176211A (en) 2021-07-27
CN113176211B CN113176211B (en) 2023-03-14

Family

ID=76926409

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110458849.3A Active CN113176211B (en) 2021-04-27 2021-04-27 Day cloud detection method based on HSV color space model transformation

Country Status (1)

Country Link
CN (1) CN113176211B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114140697A (en) * 2021-09-02 2022-03-04 广东海启星海洋科技有限公司 Surface flow field remote sensing detection method and device
CN114235148A (en) * 2022-02-25 2022-03-25 南京信息工程大学 Road night illumination quality monitoring method based on noctilucent remote sensing data
CN115690502A (en) * 2022-11-02 2023-02-03 珠江水利委员会珠江水利科学研究院 Method and system for eliminating water ripples of near-shore and inland water bodies and readable storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102637301A (en) * 2012-03-20 2012-08-15 武汉大学 Method for automatically evaluating color quality of image during aerial photography in real time
CN103439757A (en) * 2013-09-10 2013-12-11 海全胜 Cloud detection method by using MODIS remote sensing thermal infrared data
US20190064055A1 (en) * 2017-08-22 2019-02-28 Harris Corporation Intelligent Sensor Pointing for Remote Sensing Applications
CN109767465A (en) * 2018-04-23 2019-05-17 中南大学 A method of the mist rapidly extracting on daytime based on H8/AHI
US20200250428A1 (en) * 2019-02-04 2020-08-06 Farmers Edge Inc. Shadow and cloud masking for remote sensing images in agriculture applications using a multilayer perceptron
CN112001374A (en) * 2020-10-28 2020-11-27 航天宏图信息技术股份有限公司 Cloud detection method and device for hyperspectral image

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102637301A (en) * 2012-03-20 2012-08-15 武汉大学 Method for automatically evaluating color quality of image during aerial photography in real time
CN103439757A (en) * 2013-09-10 2013-12-11 海全胜 Cloud detection method by using MODIS remote sensing thermal infrared data
US20190064055A1 (en) * 2017-08-22 2019-02-28 Harris Corporation Intelligent Sensor Pointing for Remote Sensing Applications
CN109767465A (en) * 2018-04-23 2019-05-17 中南大学 A method of the mist rapidly extracting on daytime based on H8/AHI
US20200250428A1 (en) * 2019-02-04 2020-08-06 Farmers Edge Inc. Shadow and cloud masking for remote sensing images in agriculture applications using a multilayer perceptron
CN112001374A (en) * 2020-10-28 2020-11-27 航天宏图信息技术股份有限公司 Cloud detection method and device for hyperspectral image

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
李微 等: "基于HSV色彩空间的MODIS云检测算法研究", 《中国图像图形学报》 *
高贤君 等: "航空摄影过程中云的实时自动检测", 《光谱学与光谱分析》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114140697A (en) * 2021-09-02 2022-03-04 广东海启星海洋科技有限公司 Surface flow field remote sensing detection method and device
CN114235148A (en) * 2022-02-25 2022-03-25 南京信息工程大学 Road night illumination quality monitoring method based on noctilucent remote sensing data
CN115690502A (en) * 2022-11-02 2023-02-03 珠江水利委员会珠江水利科学研究院 Method and system for eliminating water ripples of near-shore and inland water bodies and readable storage medium
CN115690502B (en) * 2022-11-02 2023-06-13 珠江水利委员会珠江水利科学研究院 Method, system and readable storage medium for eliminating water wave of inland and coastal water body

Also Published As

Publication number Publication date
CN113176211B (en) 2023-03-14

Similar Documents

Publication Publication Date Title
CN113176211B (en) Day cloud detection method based on HSV color space model transformation
Shao et al. Cloud detection in remote sensing images based on multiscale features-convolutional neural network
Wei et al. Cloud detection for Landsat imagery by combining the random forest and superpixels extracted via energy-driven sampling segmentation approaches
CN109581372B (en) Ecological environment remote sensing monitoring method
Zhu et al. Object-based cloud and cloud shadow detection in Landsat imagery
Adeline et al. Shadow detection in very high spatial resolution aerial images: A comparative study
US7184890B2 (en) Cloud shadow detection: VNIR-SWIR
CN109376591B (en) Ship target detection method for deep learning feature and visual feature combined training
CN112200787B (en) Cloud detection method, storage medium and system for optical remote sensing image
CN103729848A (en) Hyperspectral remote sensing image small target detection method based on spectrum saliency
CN102750701A (en) Method for detecting spissatus and spissatus shadow based on Landsat thematic mapper (TM) images and Landsat enhanced thematic mapper (ETM) images
KR20080031702A (en) Method of cloud masking in remote sensing images
KR102461468B1 (en) Method for detecting a red tide using multispectral images and system thereof
WO2019184269A1 (en) Landsat 8 snow-containing image-based cloud detection method
CN107103295B (en) Optical remote sensing image cloud detection method
CN104537639A (en) Translation evaluation based optimized spectrum image data fusion method
CN112818880B (en) Aerial image vegetation extraction and classification method based on deep learning
Iannone et al. Proba-V cloud detection Round Robin: Validation results and recommendations
CN112418156B (en) Unsupervised classification-based optical remote sensing image sea surface ship target detection method
Connor et al. Discriminative imaging using a LWIR polarimeter
Zhao et al. Assessment of SPOT-6 optical remote sensing data against GF-1 using NNDiffuse image fusion algorithm
CN111695503B (en) Coral reef bottom material classification method based on dual-band radiance
Shi et al. Study on Cloud Classifications by Using AVHRR. GMS-5 and TERRA/MODIS Satellite Data
CN111340779B (en) Comprehensive detection method for quasi-circular vegetation patches
Sun et al. An Improved Fmask Algorithm in Tropical Regions for Landsat Images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant